Contents
Nearly every software product for the trades now says it has AI. Some of it saves real hours every week. Some of it is a chat window bolted onto the same product at a higher price. For a small service business, the useful question isn't "should we use AI?" It's "which jobs can AI do well enough today, and which still need a person?" This guide gives an honest answer, plus a way to test any tool before you let it talk to your customers.
What "AI" usually means in these products
"AI" covers several different technologies. Knowing which one you're being sold helps you judge it.
- Language models read and write text. They draft emails, summarize notes, answer questions and power most chatbots.
- Speech tools turn calls into text, and some voice systems answer the phone and talk back.
- Image tools describe, sort or generate pictures.
- Prediction and optimization suggest routes, schedules or prices. Some of this is older math with a new label, which is fine as long as it works.
Where AI is genuinely useful today
AI does best at tasks that are mostly about text, where a mistake is cheap and a person checks the result before it reaches a customer.
Writing first drafts. Service page outlines, Google Business Profile posts, email newsletters, job ads, review responses and seasonal offers. AI gets you from a blank page to a workable draft in a minute. You then add the real details: your towns, your warranty, the jobs you actually do.
Cleaning up notes. A tech can dictate a rambling voice memo in the truck ("swapped the flapper, fill valve was shot too, told her the wax ring looks original") and AI can turn it into a clear summary for the homeowner and a tidy note for your records.
Summarizing. A long email thread with a property manager, a call transcript, a pile of customer feedback. AI is good at pulling out the main points and the next steps.
Answering routine questions. Hours, service area, whether you work on a certain kind of system, how to book. A well-set-up assistant can handle these after hours, as long as it hands anything unusual to a person.
Sorting incoming requests. Tagging new inquiries by service and urgency, so the office sees the no-heat call before the "quote for next spring" email.
Translating. Drafting a message in a customer's preferred language, with a fluent speaker checking anything important when you can.
Office paperwork. Turning your process into a written checklist, drafting a training outline for new hires, or explaining a confusing supplier invoice.
Where AI is weak or risky today
It makes things up. Language models can state false things with complete confidence. They might invent a warranty term, a price, a license number, a code requirement or a policy you don't have. Never let AI tell customers about prices, warranties, safety or code unless it's working only from information you've given it and checked.
It can't see the job. AI can list possible causes of a noisy furnace or a tripping breaker, but it can't replace a trained person on site. Anything involving safety, such as a gas smell, a sparking outlet, a sagging ceiling or water near a panel, must go straight to a person and, where needed, to emergency services.
It guesses at quotes. Without your price book and a proper look at the job, any number it produces is a guess, and customers will hold you to it.
Phone calls are hard. Voice systems have improved, but noisy job sites, strong accents, bad connections, upset callers and real emergencies still trip them up, and some callers hang up as soon as they hear a machine. Rules on recording calls and on telling callers they're speaking with an automated system vary by state and country, so check what applies to you before switching one on.
Generic content doesn't stand out. AI-written pages tend to sound like everyone else's. Google's published guidance focuses on whether content is helpful rather than whether AI wrote it, and it treats pages mass-produced to game rankings as spam. What makes a page useful is the part AI doesn't know: your real jobs, photos, prices and local detail.
It can get numbers wrong. Check any math in a quote, invoice or budget that AI touched.
Privacy. Anything you paste into a tool may be stored by the provider. Read the data policy, and never paste card numbers, alarm codes, gate codes or anything a customer told you in confidence into a tool you haven't checked.
Fake reviews are off limits. Never use AI to write reviews of your own business or to post as customers. Review platforms ban it, and in the US the Federal Trade Commission has rules against fake reviews.
Quick reference: AI or a person?
| Task | Let AI draft it? | What a person does |
|---|---|---|
| Review responses | Yes | Check the tone, especially on negative reviews |
| Service page drafts | Yes | Add your real services, towns, photos and proof |
| Job summaries from tech notes | Yes | The tech confirms nothing was added or lost |
| Routine after-hours questions | Yes, with limits | Take over anything urgent, unusual or upset |
| Diagnosing problems | Only as a list of possibilities | Always diagnose on site |
| Quotes and prices | Only from your own price book | Always check before it's sent |
| Safety and emergencies | No | Always handle directly |
| Collections and disputes | No | Always handle directly |
How to test an AI tool before you trust it
A demo shows you the best case. Your customers will show you everything else. Before a tool talks to customers, run your own test.
- Pick one job with a clear measure. For example, "answer after-hours messages so fewer leads wait until morning" or "cut the time spent writing review responses."
- Write 20 realistic test cases from your own business. Include easy ones and hard ones: a simple booking, a price question, an angry customer, a caller who prefers Spanish, a gas smell at 2 a.m., a question about a service you don't offer, and someone trying to get the tool to promise a discount.
- Run every case and grade the answers. Was it correct? Was it safe? Did it sound like your business? Did it know when to hand off to a person?
- Watch what it does when it doesn't know. A good tool says "let me get someone for you." A bad one makes something up.
- Test the hand-off. When a person needs to take over, how do they find out, and how fast?
- Check the data. Where are recordings and transcripts stored? Who can see them? Can you delete them? Is your customers' information used to train anyone's models?
- Start with a person reviewing everything. Loosen the reins only after the tool has earned it.
Questions to ask any vendor selling AI
Examples by trade
The same rule applies everywhere: let AI handle the words around the job, and keep people on the judgment calls.
- Plumbing: AI can draft a clear explanation of repair options from a tech's notes. It should never tell a caller whether a gas smell is safe to ignore.
- HVAC: AI can write the fall tune-up reminder and sort incoming requests by urgency. It shouldn't recommend a system size without a proper load calculation by a person.
- Electrical: AI can turn a panel upgrade estimate into plain language for the homeowner. Anything about whether a situation is dangerous goes to a licensed electrician.
- Cleaning: AI can draft checklists for each type of clean and answer questions about what's included. Quotes for unusual homes still need someone to look at photos or visit.
- Landscaping: AI can help describe a design idea and draft the proposal text. Plant choices for a specific yard still depend on someone who has seen the soil, shade and drainage.
- Roofing: AI can summarize an inspection for the homeowner. It shouldn't tell anyone what their insurance will cover.
Prompts to try this week
You can test the useful side of AI today with a general chat assistant. Replace the brackets with your own details, and remove customer names and personal information first.
- Review response: "Write a short, warm reply to this Google review for my [trade] business. Thank them by first name, mention the specific job, and don't offer discounts. Keep it under 60 words. Review: [paste review]"
- Job summary: "Turn these technician notes into a short, friendly summary for the homeowner. Use plain language. Don't add anything that isn't in the notes. Notes: [paste notes]"
- Service page outline: "Outline a service page for [service] in [city]. Include sections for what's included, signs you need it, what the visit looks like, how pricing works and common questions. Leave placeholders where I need to add real details."
- Checklist: "Write a 10-step arrival checklist for a technician on a [type of call]." Have your most experienced tech review it before anyone uses it.
- Seasonal message: "Write a two-sentence reminder for past customers about [service] before [season]. Friendly, no pressure, with one line about how to book."
Our 100 AI prompts for home service businesses has many more.
What's likely to stay true
Nobody can say exactly where AI will be in a few years, and it's wise to be wary of anyone who claims to. A few things are likely to hold:
- Customers still want a person they trust in their home. AI can speed up the paperwork around a job. It doesn't replace showing up on time and doing good work.
- Accurate information about your business matters more. As more people ask AI assistants for recommendations, those assistants draw on what's published about you: your website, your listings and your reviews. Clear, consistent, specific information gives them less to get wrong.
- The owners who get the most from AI use it for the tedious parts and keep their own judgment on anything involving safety, money or a customer's trust.
Zome's free AI report is an example of the low-risk kind: it reviews your public website, Google listing and reviews, and gives you a plain-language list of things to check. You decide what to act on.
